The traditional public switched telephone network (PSTN) consists of signaling nodes connected via dedicated signaling system 7 (SS7) signaling links. The three primary types of signaling nodes in the conventional PSTN network are service switching points (SSPs), signal transfer points (STPs), and service control points (SCPs). Service switching points are end office switches that handle both voice and data traffic. Signal transfer points are switching nodes that route SS7 messages between SS7 signaling points. Service control points are databases and associated computers that provide data in response to SS7 queries. Examples of such data include billing information, 800 number translation information, and number portability information.
These conventional SS7 nodes have typically communicated by sending SS7 messages to each other over dedicated SS7 signaling links. While such signaling links provide a highly reliable means for communicating SS7 traffic, SS7 signaling links provide only fixed bandwidth to a user regardless of the user's needs. As a result, users must install or lease sufficient call signaling bandwidth to handle peak or worst-case traffic conditions. Installing or leasing sufficient call signaling bandwidth for peak conditions is inefficient since peak conditions rarely occur. Moreover, because SS7 call signaling bandwidth is expensive, there exists a need for an alternative to dedicated SS7 signaling links.
FIG. 1 is a block diagram of the conventional PSTN network. In FIG. 1, SSPs 100 and 102 communicate with SSPs 104 and 106 through STPs 108, 110, 112, and 114. SCP database nodes 116 and 120 provide data in response to queries from SSPs 100, 102, 104 and 106 and/or from STPs 108, 110, 112, and 114. All of the lines interconnecting the nodes in FIG. 1 represent conventional SS7 signaling links. As stated above, such links are often under-utilized and are expensive to install or lease.
In packet-based networks, such as transmission control protocol/Internet protocol (TCP/IP) networks, bandwidth can be shared among multiple users. In addition, the growth and popularity of the global Internet have made components for such networks readily available and cost efficient. However, integrating the traditional PSTN network with a packet-based network, such as a TCP/IP network, creates a number of problems.
For example, one problem with sending traditional call signaling traffic over a TCP/IP network is that in a TCP/IP network, transmissions between a sender and a receiver are stream-oriented. That is, TCP software on a sending machine is not guaranteed to send data in the same boundaries defined by a sending application. The amount of data sent over a TCP connection depends on the window size advertised by the receiver, the number of bytes of data that have been acknowledged by the receiver, and the maximum segment size of the physical network connecting the sender and the receiver. Accordingly, the receiving application may not receive data in the same boundaries created by the sending application. Thus, when sending call signaling messages over a TCP/IP network, several messages may be combined in one TCP segment. Alternatively, a single call signaling message may be divided among multiple TCP segments. In conventional networks, it is the job of the receiving application to parse the incoming data stream and extract the individual packets. Such parsing is difficult and increases the complexity of application programs that utilize TCP.
Another problem with sending conventional call signaling messages over a TCP/IP connection is that the timeout period for disabling a connection in TCP is too long for call signaling applications. For example, some implementations of TCP include a keep-alive timer. The keep-alive timer is reset every time a TCP segment is received. When the timer expires, it causes one side of the connection to determine if the other side is still operating. No mechanism is specified in the TCP protocol specifications for determining whether the other side is operating. In addition, the timeout period for the keep-alive timer is on the order of minutes. Thus, one side of a connection could go down and the other side could wait for minutes before resetting the connection. Such a long timeout period wastes resources on the machine that is waiting for data from the other side and is unsuitable for telephony applications.
Yet another problem with integrating conventional telephony and packet-based networks, such as TCP/IP networks, is that TCP/IP requires lengthy handshake procedures for connection establishment and termination. For example, in order to establish a TCP connection, a client application sends a synchronization (SYN) packet to a server application. The server application then sends an acknowledgement (ACK) and a SYN back to the client. The client then sends an acknowledgement to the SYN+ACK from the server. During the initial exchange of SYN and ACK messages, the client and server exchange sequence numbers. Once the client sends acknowledgement to the SYN+ACK to the server, the TCP software on the client machine is in an open state in which data can be received from the server and data from the sending application can be sent to the server.
In order to terminate a TCP connection, when an application closes a connection, the TCP software associated with that application sends a packet having a finish (FIN) bit set, referred to as a FIN packet, to the TCP software on the other side of the connection. The TCP software of the machine that receives the FIN sends an ACK to the FIN and informs the application that a FIN has been received. If the application is finished sending data, the application closes the connection. In response to the application close, the TCP software sends a FIN to the TCP software that sent the original FIN. In response to receiving the FIN, the TCP software sends an ACK. Once this ACK is sent, the connection is considered to be closed by both sides of the connection.
While TCP connection establishment and termination procedures have proven to be reliable and account for a variety of error conditions, such procedures are cumbersome and require many round trip times in order to complete. For example, in TCP connection establishment, a minimum of 1.5 round trip times is required. In the TCP connection termination scenario described above, at least two round trip times are required. In addition, TCP software on both sides of the connection is required to maintain state and perform additional processing during connection establishment and termination.
For all of these reasons, the number of occurrences of TCP connection establishment and termination procedures should be minimized. For example, if it is desired to upgrade software in a telephony device that currently communicates with a remote device over a TCP connection, the connection must be terminated. Connection termination requires the handshaking procedure discussed above. Once the software is upgraded, the connection must be reestablished. Connection reestablishment requires the three-way handshaking procedure described above. Thus, performing a software upgrade requires an initial TCP connection establishment, a TCP connection termination, followed by another TCP connection establishment. These procedures waste resources and should be minimized, especially in high-traffic telecommunications switches.
In light of all these difficulties associated with integrating conventional telephony networks, such as SS7 networks, and stream-oriented packet-based networks, such as TCP networks, there exists a need for novel methods and systems for integrating these networks that avoid at least some of the difficulties associated with the prior art.